A study of fire severity across three spatiotemporal scales using remote sensing, environmental factors and plant ecology

The study of fire severity (overall ecological damage produced by a fire) is of increasing importance under climate change. Climate change will impact fire regimes by altering fire frequencies, mediated by changes in fire weather, moisture contents and plant productivity. A consequence of these changing fire regimes will be more frequent and severe burning of fire‚ÄövÑv™adapted and fire‚ÄövÑv™sensitive ecosystems, likely affecting their ability to recover. My thesis employs three spatiotemporal scales to identify fire severity drivers and assess plant responses to fire using remote sensing, environmental factors and plant traits. In the general discussion, I combine all findings from the three scales into globally relevant results, and discuss research gaps for future investigation. After an introductory chapter, I explore large‚ÄövÑv™scale patterns of fire severity. For this, I use remote sensing (satellite images), Random Forests and generalised additive modelling to: (a) identify the main environmental factors driving fire severity of 127 wildfires and prescribed burns across forested and treeless vegetation, over 425,000 hectares in western Tasmania, for a 25‚ÄövÑv™year period; and (b) validate through remote sensing that forest wildfires have greater fire severity than prescribed burns. Western Tasmania is a challenging region for multi‚ÄövÑv™temporal remote sensing analysis due to frequent cloud cover. Given the presence of large treeless communities in western Tasmania, which recover quickly after fire, I ensured that post‚ÄövÑv™fire satellite images were as close as possible in time to the fire event, and both pre‚ÄövÑv™fire and post‚ÄövÑv™fire images of each fire spanned the same time of the year, diminishing seasonal and phenological effects. Through this approach, I obtained stronger relationships between satellite images and environmental factors in treeless vegetation than in forests. My results showed that annual precipitation in forests (negative relationship), and summer precipitation in forested and treeless vegetation (positive relationship), were the main drivers of remotely‚ÄövÑv™sensed fire severity, reflecting the link between plant productivity and water availability. I also confirmed that remotely‚ÄövÑv™sensed fire severity of forest wildfires was statistically higher than that of prescribed burns. However, there was no statistical difference in remotely‚ÄövÑv™sensed fire severity in treeless vegetation burnt by wildfires and prescribed burns, due to the fewer vegetation layers present in treeless vegetation available to burn, relative to forests. In the third chapter, based on a medium spatiotemporal scale, I assess the responses to fire of Eucalyptus delegatensis subsp. tasmaniensis. The study area, encompassing 25,000 hectares in Central Tasmania, was burnt at different severities over two months in 2016 by lightning fires, after a protracted drought, impacting large tracts of E. delegatensis forests with a rich history of logging. Here, I employ plant traits to assess post‚ÄövÑv™fire recovery strategies of this subspecies and compare these strategies with those of its mainland Australian counterpart (subsp. delegatensis). My results indicated that the Tasmanian subspecies recovers primarily by epicormic resprouting after fire, contrarily to the obligate‚ÄövÑv™seeder mainland subspecies. Epicormic resprouting was positively associated with fire severity and stem diameter, but was not affected by logging intensity. By contrast, tree survival was reduced as fire severity and logging intensity increased, with larger trees (diameter ‚Äöv¢‚Ä¢ 20 cm) tending to survive more frequently. Seedling recruitment was limited, presumably due to protracted drought before the fires and extremely wet conditions soon after the fires. On the whole, heavily‚ÄövÑv™logged, burnt forests of Tasmanian E. delegatensis will be at higher risk of burning again in the next three decades because of climate change, since all the regrowth will be equally vulnerable to fire kill. In the fourth chapter (small spatiotemporal scale), I use plant traits to assess the postfire recovery of a Brazilian savanna‚ÄövÑv™type (1.5 hectares) burnt under prescribed conditions in south‚ÄövÑv™east Brazil. I investigate the role of plant traits in protecting savanna trees from fire, and the relationship between these traits and the two main post‚ÄövÑv™fire recovery strategies. My analyses suggested that tree height is the most advantageous trait to protect savanna trees from fire. Although bark thickness protects vascular cambium, tall trees experience less damage by the flames and heat than small trees, whose aerial parts are near the ground. In addition, I found alternative recovery strategies related to height: tall trees (height ‚Äöv¢‚Ä¢ 3.7 m) tended to recover predominantly through epicormic resprouts, whereas short trees (diameter ‚Äöv¢¬ß 5 cm) recovered mostly via basal resprouts. This differentiation may relate to carbohydrate allocation, with trees that grow high investing in epicormic buds, so that they can resprout quickly from the stem after fire, while short trees storing buds in the base of the tree to allow for basal resprouting. Furthermore, I consider the possibility of the cerrado falling into the fire trap due to frequent burning of regenerating cerrado communities because of climate change. The final chapter considers the overall implications of this work. I found a connection between environmental factors and plant communities that favours the development of plant traits to survive fire. However, climate change can break this connection by transforming fire regimes. Changes in precipitation and temperature will affect plant communities adapted to specific fire regimes and fire‚ÄövÑv™sensitive communities that require long fire‚ÄövÑv™free periods. Tasmanian E. delegatensis and Brazilian savannas, which resprout after fire, may experience resprouting failure if fires occur more frequently and fire intensities increase. For example, resprouting communities of Mediterranean, mixed‚ÄövÑv™eucalypt and boreal forests, subject to varying fire regimes, have perished or undergone interval squeeze. My thesis provides evidence that remotely‚ÄövÑv™sensed data can be successfully employed to study fire severity in complex landscapes, including treeless vegetation and prescribed burns. The increasing availability of remote sensing technologies will offset previous obstacles related to meteorological phenomena, seasonal effects and low revisiting times. New remote sensing technologies and artificial intelligence will allow monitoring fire‚ÄövÑv™adapted plant communities that are repeatedly burnt, and fire‚ÄövÑv™sensitive plant communities susceptible to interval squeeze, to continue adding knowledge of the impacts of climate change.